I think this misses the point. I’m sure it’s not the Review you would have written. It has many flaws in any case. It’s still remarkable - 6 minutes!
It would have been unthinkable that it could do this when the review launched. It’s time to look at the rate and direction of progress and plan accordingly.
Despite DeepResearch’s capabilities, I’m still stuck on the issue of trust.
Is there a difference between the DeepResearch output here and the output of, say, a Keith Dear who took the time to digest all the relevant documents and generate a new Defence Review? I think there is, if for no other reason that by doing that research you become a trustworthy source of the output.
You, after all, are qualified to evaluate this DeepResearch output precisely because you’ve spent years thinking (and reading) about these issues.
If we—slowly at first—start to turn these tasks over to AI, won’t we, at the very same time, cause ourselves to become people who are no longer qualified to judge their output?
By asking the machine to do the work for us, do we lose the expertise we need to judge the machines’ performance?
Hi. I write these comments as a former infantry C/Sgt who left the army in 2004 and to that end have seen some parts of previous SDR’s but not a whole document. And to be honest being at the lowest point of any impact from it i don’t think I needed to. The document is very bold and innovative and has given me glimpse of future warfare that I could never imagine, to be written in six minutes is incredible and I truly believe some of it should be adopted now, especially the procurement and development part. Ours is, and always has been abysmal, from the SA80 to Ajax. My two main observations on the document are relatively simple but I believe very important. Whilst it discusses long range precision munitions, swarm attacks, drone re supply, inter service strike capabilities etc etc, I did not see anything on how this intelligent war fighting system is going to protect itself from hacking attack. I see quantum computing mentioned and I am sure the Chinese in particular are a long way ahead in this field than anyone else. If this system was hacked then we would be defenceless. If it worked, and that would be amazing, it’s ok to kill the enemy before he knows what he’s doing but then what? To hold ground regardless of your drone capability you will still need boots on the ground. Thanks for sharing, it’s something to be very proud, and a little scared of.
Lewis - thanks for reading, and even more so for taking the time to comment. I think the protection from hacking is probably covered by this:
'Cyber defense and electronic warfare improvements also come under this umbrella: our networks must resist jamming and hacking. In exercises, we will practice “fighting fractured” – assuming networks are contested – to refine the tech and tactics for staying linked under fire. A resilient networked force can coordinate kill chains even in chaos, whereas an adversary with disrupted comms will be paralyzed.'
But I agree it is high level, and could do with elaboration.
I'm not with you on needing boots on the ground in order to defend it. Sure, that's true today, but it isn't a given that it remains so. Increasingly advanced autonomous systems might well see the battlefield come to resemble modern chess competitions. Where humans in the human-machine team are the limiting factor on performance. The 'wetware' of our brains holding back the software and hardware of machine intelligence.
I share this stuff because I am scared by it. I don't think we are taking AI even a 100th as seriously as we should be, and the consequences - economic, security, social - will be at best disastrous and at worst catastrophic.
Is it a Defence Review or a pastiche of a Defence Review though Keith?
I think this misses the point. I’m sure it’s not the Review you would have written. It has many flaws in any case. It’s still remarkable - 6 minutes!
It would have been unthinkable that it could do this when the review launched. It’s time to look at the rate and direction of progress and plan accordingly.
This is fascinating, Keith. Thanks for sharing.
Despite DeepResearch’s capabilities, I’m still stuck on the issue of trust.
Is there a difference between the DeepResearch output here and the output of, say, a Keith Dear who took the time to digest all the relevant documents and generate a new Defence Review? I think there is, if for no other reason that by doing that research you become a trustworthy source of the output.
You, after all, are qualified to evaluate this DeepResearch output precisely because you’ve spent years thinking (and reading) about these issues.
If we—slowly at first—start to turn these tasks over to AI, won’t we, at the very same time, cause ourselves to become people who are no longer qualified to judge their output?
By asking the machine to do the work for us, do we lose the expertise we need to judge the machines’ performance?
Maybe. I think the honest answer is that we don't know. So much must now change.
Hi. I write these comments as a former infantry C/Sgt who left the army in 2004 and to that end have seen some parts of previous SDR’s but not a whole document. And to be honest being at the lowest point of any impact from it i don’t think I needed to. The document is very bold and innovative and has given me glimpse of future warfare that I could never imagine, to be written in six minutes is incredible and I truly believe some of it should be adopted now, especially the procurement and development part. Ours is, and always has been abysmal, from the SA80 to Ajax. My two main observations on the document are relatively simple but I believe very important. Whilst it discusses long range precision munitions, swarm attacks, drone re supply, inter service strike capabilities etc etc, I did not see anything on how this intelligent war fighting system is going to protect itself from hacking attack. I see quantum computing mentioned and I am sure the Chinese in particular are a long way ahead in this field than anyone else. If this system was hacked then we would be defenceless. If it worked, and that would be amazing, it’s ok to kill the enemy before he knows what he’s doing but then what? To hold ground regardless of your drone capability you will still need boots on the ground. Thanks for sharing, it’s something to be very proud, and a little scared of.
Lewis - thanks for reading, and even more so for taking the time to comment. I think the protection from hacking is probably covered by this:
'Cyber defense and electronic warfare improvements also come under this umbrella: our networks must resist jamming and hacking. In exercises, we will practice “fighting fractured” – assuming networks are contested – to refine the tech and tactics for staying linked under fire. A resilient networked force can coordinate kill chains even in chaos, whereas an adversary with disrupted comms will be paralyzed.'
But I agree it is high level, and could do with elaboration.
I'm not with you on needing boots on the ground in order to defend it. Sure, that's true today, but it isn't a given that it remains so. Increasingly advanced autonomous systems might well see the battlefield come to resemble modern chess competitions. Where humans in the human-machine team are the limiting factor on performance. The 'wetware' of our brains holding back the software and hardware of machine intelligence.
I share this stuff because I am scared by it. I don't think we are taking AI even a 100th as seriously as we should be, and the consequences - economic, security, social - will be at best disastrous and at worst catastrophic.